sessions

$npx mdskill add microsoft/vscode/sessions

Use this skill for implementation, review, or design work under `src/vs/sessions/**`.

SKILL.md

.github/skills/sessionsView on GitHub ↗
---
name: sessions
description: Core principles and workflow router for changes to the Agents Window under src/vs/sessions.
---

# Agents Window development

Use this skill for implementation, review, or design work under `src/vs/sessions/**`.

## 1. Apply the core principles

- Preserve the layer direction: `vs/sessions` may import `vs/workbench` and lower layers; `vs/workbench` must never import `vs/sessions`.
- Keep shared Sessions code provider-neutral. Non-provider contributions must not import provider implementations.
- Model mutable session and chat state with observables. Use events for notifications, not as a parallel state model or for control flow.
- Register Sessions menu IDs in `browser/menus.ts` and consume `Menus.*`.
- Import contributions from the appropriate `sessions.*.main.ts` entry point.
- Prefer Sessions-owned adaptations over shared workbench changes unless the capability is genuinely shared.
- Put stable architecture in the owning specification and concrete behavior in tests. Do not preserve implementation chronology as development guidance.

## 2. Identify the owning area

Start with `src/vs/sessions/README.md`, then read only the specifications relevant to the change:

| Area | Specification |
|------|---------------|
| Layering, folder ownership, cross-module imports | `src/vs/sessions/LAYERS.md` |
| Session/chat model, services, provider contract, core data flow | `src/vs/sessions/SESSIONS.md` |
| Automations ownership, routing, migration, persistence, and run lifecycle | `src/vs/sessions/AUTOMATIONS.md` |
| Workbench parts, grid, title bar, editor presentation | `src/vs/sessions/LAYOUT.md` |
| Session-aware layout state and restoration | `src/vs/sessions/LAYOUT_CONTROLLER.md` |
| Single-pane behavior and expected compositions | `src/vs/sessions/SINGLE_PANE_SCENARIOS.md` |
| Sessions sidebar list, grouping, filtering, and persistence | `src/vs/sessions/SESSIONS_LIST.md` |
| Phone layout and mobile components | `src/vs/sessions/MOBILE.md` |
| AI customizations | `src/vs/sessions/AI_CUSTOMIZATIONS.md` |
| Copilot customizations | `src/vs/sessions/copilot-customizations-spec.md` |
| Copilot Chat provider | `src/vs/sessions/contrib/providers/copilotChatSessions/COPILOT_CHAT_SESSIONS_PROVIDER.md` |
| Agent Host provider | `src/vs/sessions/contrib/providers/agentHost/AGENT_HOST_SESSIONS_PROVIDER.md` |
| Remote Agent Host provider | `src/vs/sessions/contrib/providers/remoteAgentHost/REMOTE_AGENT_HOST_SESSIONS_PROVIDER.md` |

Do not load the learning inbox by default. Search its headings and scopes, then read only matching entries after the authoritative specification.

## 3. Inspect before changing

- Trace the current implementation and its existing tests.
- Search for shared helpers, context keys, menu IDs, entry-point imports, and provider abstractions before adding new ones.
- Confirm which layer owns the behavior. Keep provider-specific decisions in the provider and view/layout decisions in Sessions-owned browser code.
- For UI work, also invoke the applicable accessibility, design, CSS, layout, or theming skill.
- For agent, LLM, policy, permissions, telemetry, or managed-setting changes, invoke the applicable specialist skill before implementation.

## 4. Implement the contract

Apply the core principles and the focused specification. Prefer small changes that preserve these boundaries:

- `ISessionsManagementService` owns model orchestration and provider routing.
- `ISessionsService` owns visible and active session behavior.
- Providers expose provider-neutral state through `ISession` and `IChat`.
- Session state is observable; consumers derive UI state reactively.
- Contributions load through the appropriate `sessions.*.main.ts` entry point.
- Sessions menus use the shared `Menus` registry.
- Shared workbench changes represent shared capability, not Sessions-specific policy.

### Specification edit gate

Bug fixes do not update specifications when they restore an existing contract. Before editing an authoritative specification, identify all three:

1. the existing ownership, interface, lifecycle, state-machine, persistence, or cross-component contract that intentionally changes;
2. the implementation surfaces affected by that contract change;
3. why a regression test and a brief code comment cannot fully represent it.

If any answer is missing, leave the specification unchanged. Put concrete behavior in a focused test, keep a non-obvious implementation constraint beside the owning code, and preserve investigation history in the issue or pull request.

Update a specification only when component ownership, an interface or lifecycle contract, a state machine, persistence, or a cross-component invariant changes. Do not update specifications for styling, copy, action placement, telemetry fields, settings defaults, implementation algorithms, or individual bug fixes. Those details belong in code and focused tests.

## 5. Validate proportionally

Run the smallest existing checks that cover the change:

- focused unit tests for affected behavior;
- `npm run valid-layers-check` when imports or module ownership change;
- targeted type checking or compilation when TypeScript changes warrant it;
- relevant integration, E2E, or visual validation for cross-process or UI work.

Documentation-only changes require link, path, and consistency checks rather than a full build.

## 6. Record feedback correctly

When a user explicitly corrects or rejects an approach, invoke the `feedback-learning` skill unless they use the literal `learn!` trigger. Literal `learn!` requests follow `.github/instructions/learnings.instructions.md` instead. A durable architecture invariant belongs in the owning specification, concrete behavior belongs in a regression test, and unproven reusable guidance belongs temporarily in the scoped learning inbox. Never append every correction to this skill.

## 7. Maintain this skill

Update this skill only when a principle is stable, cross-cutting, and useful for most Agents Window work, or when the routing/workflow itself changes. Put subsystem contracts in their focused specification and bug behavior in tests.

Keep the core-principles section at no more than ten bullets. Before adding one, merge overlap, remove obsolete guidance, and prefer rewriting an existing principle. Never append incident-specific details or use this skill as a learning log.

More from microsoft/vscode

SkillDescription
accessibilityPrimary accessibility skill for VS Code. REQUIRED for new feature and contribution work, and also applies to updates of existing UI. Covers accessibility help dialogs, accessible views, verbosity settings, signals, ARIA announcements, keyboard navigation, and ARIA labels/roles.
act-on-feedbackAct on user feedback attached to the current session. Use when the user submits feedback on the session's changes via the Submit Feedback button.
agent-customization**WORKFLOW SKILL** — Create, update, review, fix, or debug VS Code agent customization files (.instructions.md, .prompt.md, .agent.md, SKILL.md, copilot-instructions.md, AGENTS.md). USE FOR: saving coding preferences; troubleshooting why instructions/skills/agents are ignored or not invoked; configuring applyTo patterns; defining tool restrictions; creating custom agent modes or specialized workflows; packaging domain knowledge; fixing YAML frontmatter syntax. DO NOT USE FOR: general coding questions (use default agent); runtime debugging or error diagnosis; MCP server configuration (use MCP docs directly); VS Code extension development. INVOKES: file system tools (read/write customization files), ask-questions tool (interview user for requirements), subagents for codebase exploration. FOR SINGLE OPERATIONS: For quick YAML frontmatter fixes or creating a single file from a known pattern, edit the file directly — no skill needed.
agent-host-chat-contributionsBuild and review cross-cutting agent-host chat behavior through lifecycle contributions. Use when adding turn lifecycle side effects, prompt or context injection, restored-history transformation, protocol-action observation, or when reviewing changes that add code to AgentSideEffects or AgentService.
agent-host-e2e-testsUse when writing, recording, updating, validating, or troubleshooting Agent Host end-to-end and real-provider integration tests under src/vs/platform/agentHost/test/node/e2e and providerIntegration. Covers adding a cross-provider test, re-recording fixtures after an SDK bump, required cross-platform Azure validation, gating non-deterministic or platform-specific tests, and diagnosing replay cache misses.
agent-host-logsAnalyze Agent Host debug log exports. Use when given an ah-logs or ahp-logs zip/folder, an Export Agent Host Debug Logs bundle, events.jsonl, AHP JSONL transport logs, Agent Host.log, remote-agenthost.log, or copilot-logs.
author-contributionsIdentify all files a specific author contributed to on a branch vs its upstream, tracing code through renames. Use when asked who edited what, what code an author contributed, or to audit authorship before a merge. This skill should be run as a subagent — it performs many git operations and returns a concise table.
auto-perf-optimizeRun agent-driven VS Code performance or memory investigations. Use when asked to launch Code OSS, automate a VS Code scenario, run the Chat memory smoke runner, capture renderer heap snapshots, take workflow screenshots, compare run summaries, or drive a repeatable scenario before heap-snapshot analysis.
azure-pipelinesUse when validating Azure DevOps pipeline changes for the VS Code build. Covers queueing builds, checking build status, viewing logs, and iterating on pipeline YAML changes without waiting for full CI runs.
build-champ-triageFind out why a build has failed